OpenAI Stopped Being an AI Lab. GPT-5.5 Is Just the Latest Proof.

OpenAI shipped GPT-5.5 last week and the coverage, predictably, was all about the model. Benchmarks, pricing, what it’s better at than GPT-5.4. Sure, fine. But I keep staring at the actual story here, which has nothing to do with tokens per second: OpenAI is $25 billion in annual revenue, bought a Silicon Valley talk show for hundreds of millions of dollars, raised the largest private funding round in human history just last month, and is quietly walking toward an IPO. At some point you have to ask — what exactly is this company now?

GPT-5.5 is genuinely interesting as a model. It’s the first fully retrained base model since GPT-4.5, and it’s built from the ground up to be agentic — meaning it doesn’t just answer questions, it works through problems. It can write code, browse the web, operate software, and produce documents inside a single conversation without you having to hand-hold it across tools. It comes in two tiers. The standard version is priced for developers building on the API. GPT-5.5 Pro is the enterprise-grade option — roughly six times the cost of standard, aimed at organizations running it at scale. The message is clear: this is the model for people who need AI to actually do work, not just answer questions.

It also landed six weeks after GPT-5.4. That pace used to feel shocking; now it just feels like the new normal. The model is good. The model will probably be surpassed by GPT-5.6 before summer. That’s the world we’re in.

Here’s what I’m not going to spend a lot of time on: whether GPT-5.5 beats Gemini 3.1 Pro on benchmarks. (Google is currently leading there, and DeepSeek V4 dropped the same day, because of course it did.) The model race is real, it’s competitive, and it’s also rapidly becoming a commodity problem. Everyone’s closing the gap. Which is exactly why the more interesting OpenAI story isn’t about the model at all.

Let’s talk about the company for a second.

OpenAI went from $1 billion in annualized revenue in early 2023 to $25 billion today. Three years. That’s not a trajectory — that’s a vertical line. For context: it took Salesforce about 23 years to hit $25 billion in revenue, reaching that mark around fiscal year 2022. Netflix got there in roughly the same timeframe. OpenAI is doing it in a product cycle.

They now have 900 million weekly active ChatGPT users, 50 million paying subscribers, and 9 million paying business accounts. Enterprise revenue is over 40% of the total and climbing. Last month, they closed a $122 billion funding round — the single largest private fundraise in history — at an $852 billion valuation. And they’re building toward an IPO, potentially before the end of 2026.

The TBPN acquisition is where this gets really interesting to me. In early April, OpenAI bought TBPN — Technology Business Programming Network — a daily live tech talk show hosted by founders Jordi Hays and John Coogan. It’s basically a Silicon Valley business show that runs three hours a day, five days a week. It had about $5 million in ad revenue in 2025 and was on track for $30 million this year. OpenAI paid “low hundreds of millions” for it.

Why? That’s the question that’s been bugging me since the deal dropped. OpenAI’s own explanation had something to do with creating space for “real, constructive conversation” about AI’s impact on society. The acquisition reports to Chris Lehane, OpenAI’s chief political operative — which is a tell. This isn’t about content. It’s about narrative infrastructure. It’s about owning a channel that reaches the exact people — founders, investors, tech operators — who shape the cultural context around AI. Anthropic doesn’t have that. Google doesn’t have that. Meta isn’t doing this.

This is what a platform company does when it’s winning. It doesn’t just build products — it builds the ecosystem around its products. It buys the conversation.

And the super app piece ties it all together. OpenAI has been quietly announcing, over the past several months, that it’s consolidating everything — ChatGPT, Codex, its Atlas browser — into a single desktop super app. One interface that does everything: codes, browses, chats, researches, builds documents. GPT-5.5 is essentially the engine designed to power that experience. The agentic design isn’t incidental. It’s the whole point.

The super app strategy is what separates companies that win a technology generation from companies that just win a technology moment. Microsoft won the PC era by owning the operating system. Google won search by owning the browser and then extending into everything adjacent. Apple won mobile by owning the platform. OpenAI is making its move for the AI layer — and it’s doing it by building the surface you live in, not just the model underneath.

That’s where the IPO math gets interesting and a little uncomfortable. Going public means OpenAI will answer to public markets for the first time. It means quarterly pressure, analyst calls, growth targets. The company was born as a nonprofit with a mission around safe AGI development. It’s been slowly but deliberately converting to a for-profit structure over the past couple of years. The IPO is the final step in that transformation — and once it happens, the pressure to grow into an $852 billion valuation is real and relentless.

I’m not saying that’s bad, exactly. A lot of important, genuinely useful technology gets built under that kind of pressure. But it does mean something when the company that positioned itself as the careful, safety-first alternative to fast-moving AI development is now the largest private AI company in history, buying media companies, building super apps, and eyeing Nasdaq.

GPT-5.5 is a really capable engine — but it’s an engine in service of something much bigger. OpenAI is building the place where AI-native work happens, and it’s acquiring everything it needs to control that environment: the model, the interface, the distribution, the narrative, and soon, the public markets.

Whether that’s exciting or unsettling probably depends on how much you trust the people running it. For now, I’ll say this: OpenAI stopped being an AI lab a while ago. They just kept the branding.